AI agent
auto-improve
GAN-style self-improvement loop for any text artifact: mutate, grade with a SEPARATE model, keep only verified wins (pairwise-judged), revert the rest. The git history is the improvement log.
- Rank
- #493 of 2,468
- Stars
- 100★ +50/wk
- On radar
- since 2026-07-07
Momentum
rate of rise, not size · recomputes hourly
54.8
24h· new7d· new
Hover the chart for daily values
Why it's ranked
auto-improve is an AI agent ranked #493 of 2,468 on the Cresting momentum radar, with a score of 54.8. It has 100 GitHub stars, +50 in the last 7 days, and has been tracked since 2026-07-07.
Every score decomposes into published factors, the same math for every tool, paid or not. Read the methodology →
| Velocity (weighted, cohort-normalized) | 0.888 |
| Recency boost | 0.052 |
| Signal decay | 0.994 |
| Corroboration | 1.000 |
| Quality gate | 1.000 |
Raw signals (30 days)
github · forks+10 in window · 11 latest · 7 snapshots
github · stars+50 in window · 100 latest · 7 snapshots
Related ai agents
- speak-human-tw「說人話」：繁體中文的去 AI 味改寫 skill。抓 38 種 AI 寫作痕跡，順手校正中國用語與半形標點，給 Claude Code / Codex / Cursor 用。#1 · 380★
- pilotfishMulti-model orchestration layer for Claude Code — the frontier model plans, cheaper models execute, verification guards quality. One-prompt install.#2 · 385★
- fable-methodThe Fable Workflow: how Claude Fable 5 worked, distilled into skills any model can run, with the eval that keeps it honest. Think / act / prove.#3 · 296★
- dittoMine your Claude Code and Codex logs into a local you.md agent profile.#4 · 134★
- guizang-material-illustration归藏的材质插画 skill：生成带字解释图、图表美化和参考辅助配图。#5 · 586★